fix: fix channel videos and playlists on searches

Channel videos are now encapsulated in a `lockupViewModel`.

There is 3 types of content that can be inside a `lockupViewModel`:

- LOCKUP_CONTENT_TYPE_VIDEO
- LOCKUP_CONTENT_TYPE_PLAYLIST
- LOCKUP_CONTENT_TYPE_PODCAST

This commit parses `LOCKUP_CONTENT_TYPE_VIDEO` and `LOCKUP_CONTENT_TYPE_PLAYLIST` types
to fix videos in channels and playlists in channels

There is another bugs to fix:

- Since `LOCKUP_CONTENT_TYPE_PODCAST` needs to be handled differently compared to `LOCKUP_CONTENT_TYPE_PLAYLIST`, podcasts of a channel do not appear in the list of podcasts (Example channel with podcasts: https://www.youtube.com/@GrantSanderson/podcasts)
- The playlists of a channel need a better parsing for it's channel hyperlink, since the channel hyperlink is not actually the channel, it's the updated date of the playlist or just "View full playlist"

PRs to fix those issues are welcome!

src/invidious/yt_backend/extractors.cr

@@ -630,13 +630,14 @@ private module Parsers
     end
   end
 
-  # Parses an InnerTube lockupViewModel into a SearchPlaylist.
+  # Parses an InnerTube lockupViewModel into a SearchPlaylist or a SearchVideo
   # Returns nil when the given object is not a lockupViewModel.
   #
   # This structure is present since November 2024 on the "podcasts" and
   # "playlists" tabs of the channel page. It is usually encapsulated in either
   # a richItemRenderer or a richGridRenderer.
   #
+  # Since 2026-05-21, now channel videos are encapsulated in a lockupViewModel.
   module LockupViewModelParser
     extend self
     include BaseParser
@@ -648,14 +649,52 @@ private module Parsers
     end
 
     private def parse_internal(item_contents, author_fallback)
-      playlist_id = item_contents["contentId"].as_s
+      content_type = item_contents["contentType"].as_s
 
+      if content_type == "LOCKUP_CONTENT_TYPE_VIDEO"
         thumbnail_view_model = item_contents.dig(
           "contentImage", "thumbnailViewModel"
         )
-
         thumbnail = thumbnail_view_model.dig("image", "sources", 0, "url").as_s
 
+        video_id = item_contents["contentId"].as_s
+
+        metadata = item_contents.dig("metadata", "lockupMetadataViewModel")
+        title = metadata.dig("title", "content").as_s
+        # Contains the views of the video and the published time of the video.
+        metadata_parts = metadata.dig("metadata", "contentMetadataViewModel", "metadataRows", 0, "metadataParts")
+
+        view_count_text = metadata_parts.dig(0, "text", "content").as_s
+        view_count = short_text_to_number(view_count_text || "0")
+        published = metadata_parts.dig?(1, "text", "content").try { |t| decode_date(t.as_s) } || Time.local
+
+        author_verified = has_verified_badge?(item_contents["ownerBadges"]?)
+
+        return SearchVideo.new({
+          title:              title,
+          id:                 video_id,
+          author:             author_fallback.name,
+          ucid:               author_fallback.id,
+          published:          published,
+          views:              view_count,
+          description_html:   "",
+          length_seconds:     0,
+          premiere_timestamp: Time.unix(0),
+          author_verified:    false,
+          author_thumbnail:   nil,
+          badges:             VideoBadges::None,
+        })
+      elsif content_type == "LOCKUP_CONTENT_TYPE_PODCAST"
+        # TODO
+      else # If it's a podcast, it's content_type would be "LOCKUP_CONTENT_TYPE_PODCAST"
+        thumbnail_view_model = item_contents.dig(
+          "contentImage", "collectionThumbnailViewModel",
+          "primaryThumbnail", "thumbnailViewModel"
+        )
+        thumbnail = thumbnail_view_model.dig("image", "sources", 0, "url").as_s
+
+        playlist_id = item_contents["contentId"].as_s
+
         # This complicated sequences tries to extract the following data structure:
         # "overlays": [{
         #   "thumbnailOverlayBadgeViewModel": {
@@ -681,6 +720,17 @@ private module Parsers
         metadata = item_contents.dig("metadata", "lockupMetadataViewModel")
         title = metadata.dig("title", "content").as_s
 
+        metadata_parts = metadata.dig("metadata", "contentMetadataViewModel", "metadataRows", 0, "metadataParts", 0)
+
+        if author_info = metadata_parts["text"]?
+          author = author_info["content"].as_s
+          author_id = author_info.dig?("commandRuns", 0, "onTap", "innertubeCommand", "browseEndpoint", "browseId")
+            .try &.as_s || author_fallback.id
+        else
+          author = author_fallback.name
+          author_id = author_fallback.id
+        end
+
         # TODO: Retrieve "updated" info from metadata parts
         # rows = metadata.dig("metadata", "contentMetadataViewModel", "metadataRows").as_a
         # parts_text = rows.map(&.dig?("metadataParts", "text", "content").try &.as_s)
@@ -693,14 +743,15 @@ private module Parsers
         return SearchPlaylist.new({
           title:           title,
           id:              playlist_id,
-        author:          author_fallback.name,
+          author:          author,
-        ucid:            author_fallback.id,
+          ucid:            author_id,
           video_count:     video_count || -1,
           videos:          [] of SearchPlaylistVideo,
           thumbnail:       thumbnail,
           author_verified: false,
         })
       end
+    end
 
     def self.parser_name
       return {{@type.name}}